Dimension Zooming

Brushing is an important localization operation. In a large dataset, the elements falling within a brushed subspace may be sizeable. This subset of brushed elements may itself possess interesting characteristics.

Problem
In parallel coordinates, the brushed subspace appears as a confined strip across the coordinate axes.
For a narrow brush, it may be difficult to examine the data within this confined strip.

To be able to study elements within a subspace, it is essential that we treat this subset of elements as a set of data in its own right, and place them in full view so that they can be examined as appropriate.

Solution
We introduce a magnification or distortion operation: dimension zooming.

How we do it ?
We scale up each of the dimensions with respect to the extents of the brushed subspace. The subset of elements may then be examined as an independent set. This zooming operation may be performed as many times as desired. For a large dataset occupying a large range of values, this operation may be invaluable for examining localized trends.

Figure 7 : This shows a magnified view of the brushed region and an accompanying mini-map that captures the location of the brushed region in the original space.

One common problem with such scaling up operations is that it is easy to lose context of the big picture, and be lost wandering in some isolated subspace. To maintain contextual information, we display a mini-map showing the position of the currently zoomed space in relation to the entire data space. Figure 7 shows an instance of this zooming operation and the accompanying mini-map.